AI Agent Tool Calling - Automated Dependency Update System

A multi-agent Python system that uses LangChain's tool calling pattern to automatically update dependencies with intelligent testing and rollback capabilities. It analyzes repositories, updates dependencies, tests the changes, rolls back breaking updates, and creates Pull Requests or Issues automatically.

🌟 Features

Core Capabilities

🤖 Fully Automated Updates: End-to-end automation from analysis to PR creation
🧪 Intelligent Testing: Automatically runs build/test commands to verify updates
🔙 Smart Rollback: Identifies breaking changes and rolls back only problematic major updates
✅ Auto PR Creation: Creates GitHub Pull Requests with successful updates
🔴 Auto Issue Creation: Creates GitHub Issues when updates can't be applied safely
📊 Multi-Agent Architecture: Orchestrator pattern with specialized sub-agents
🧠 AI-Powered Analysis: Uses Claude to analyze errors and identify problematic dependencies

Language Support

Detects and updates dependencies for:

JavaScript/Node.js (npm, yarn, pnpm)
Python (pip, pipenv, poetry)
Rust (cargo)
Ruby (bundler)
Java (Maven, Gradle)
PHP (Composer)
Go (go modules)

Smart Features

Automatic Build Detection: Detects how to build, test, and verify your project
Error Analysis: AI-powered parsing of error messages to identify culprits
Iterative Rollback: Tries to salvage as many updates as possible
Version Categorization: Categorizes updates as major/minor/patch
Comprehensive Reporting: Detailed PR descriptions with what was updated and why

🏗️ Architecture

This project implements a multi-agent system following the LangChain Tool Calling pattern:

Agent Hierarchy

auto_update_dependencies.py (Main Orchestrator)
├── dependency_analyzer.py (Analysis Agent)
│   └── Tools: clone, detect, check outdated
├── smart_dependency_updater.py (Smart Update Agent)
│   ├── Tools: detect build, test, write files, git ops
│   └── Sub-tools: apply updates, rollback, parse errors
└── dependency_operations.py (Helper Tools)
    └── Tools: categorize, version lookup, error analysis

Complete Workflow Diagram

┌─────────────────────────────────────────────────────────────┐
│  USER INPUT: Repository URL                                 │
└─────────────────────────┬───────────────────────────────────┘
                          │
                          ▼
┌─────────────────────────────────────────────────────────────┐
│  STEP 1: ANALYZE REPOSITORY                                 │
│  • Clone repository                                          │
│  • Detect package manager                                    │
│  • Find outdated dependencies                                │
│  • Categorize: major/minor/patch                             │
└─────────────────────────┬───────────────────────────────────┘
                          │
                          ▼
┌─────────────────────────────────────────────────────────────┐
│  STEP 2: APPLY ALL UPDATES                                  │
│  • Update ALL dependencies to latest                         │
│  • Including major version updates                           │
│  • Write updated dependency files                            │
└─────────────────────────┬───────────────────────────────────┘
                          │
                          ▼
┌─────────────────────────────────────────────────────────────┐
│  STEP 3: TEST UPDATES                                       │
│  • Run install command                                       │
│  • Run build command                                         │
│  • Run test command                                          │
│  • Capture all output                                        │
└─────────────────────────┬───────────────────────────────────┘
                          │
                ┌─────────┴─────────┐
                │                   │
         Tests Pass?         Tests Fail?
                │                   │
                ▼                   ▼
       ┌────────────────┐  ┌────────────────────────────┐
       │ CREATE PR      │  │ ANALYZE ERROR              │
       │                │  │ • Use AI to parse errors   │
       │ • Git branch   │  │ • Identify problematic pkg │
       │ • Commit       │  └───────────┬────────────────┘
       │ • Push         │              │
       │ • gh pr create │              ▼
       └────────┬───────┘  ┌────────────────────────────┐
                │          │ ROLLBACK MAJOR UPDATE      │
                │          │ • Find latest in major ver │
                │          │ • Update dependency file   │
                │          │ • Write file               │
                │          └───────────┬────────────────┘
                │                      │
                │                      ▼
                │          ┌────────────────────────────┐
                │          │ TEST AGAIN (Max 3x)        │
                │          └───┬──────────────────┬─────┘
                │              │                  │
                │         Now Pass?          Still Fail?
                │              │                  │
                │              └──────┐    ┌──────┘
                │                     │    │
                ▼                     ▼    ▼
       ┌────────────────┐  ┌──────────────────────────┐
       │ SUCCESS!       │  │ CREATE ISSUE             │
       │ PR Created     │  │ • Document what failed   │
       │                │  │ • Include error logs     │
       │ Output:        │  │ • Tag: dependencies      │
       │ • PR URL       │  │                          │
       │ • Summary      │  │ Output:                  │
       │ • Rollbacks    │  │ • Issue URL              │
       └────────────────┘  │ • Failure details        │
                           └──────────────────────────┘

1. Auto Update Orchestrator (Main Entry Point)

Main coordinator (auto_update_dependencies.py) that manages the complete workflow:

Receives repository URL or name
Checks prerequisites (Docker, GitHub token)
Orchestrates analysis and update agents
Manages end-to-end automated updates with PR/Issue creation

Functions:

analyze_repository() - Invokes the analyzer agent
smart_update_and_test() - Invokes the smart updater agent
Docker and GitHub token validation

2. Dependency Analyzer Agent (Analysis)

Specialized in finding outdated dependencies (dependency_analyzer.py):

Clones repositories
Detects package managers
Identifies outdated packages
Returns structured analysis reports

Tools:

clone_repository - Clones git repos to temp directories
detect_package_manager - Identifies package managers and config files
read_dependency_file - Reads dependency configuration files
check_npm_outdated - Checks outdated npm packages
check_pip_outdated - Checks outdated Python packages (via PyPI API)
cleanup_repository - Removes temporary files

3. Smart Dependency Updater Agent (Update & Test)

Specialized in updating with intelligent testing and rollback (smart_dependency_updater.py):

Applies dependency updates
Runs build and test commands
Automatically rolls back breaking changes
Creates GitHub PRs on success
Creates GitHub Issues on failure

Tools:

detect_build_command - Auto-detects build/test commands
apply_updates - Updates dependency files
test_updates - Runs build/test commands
rollback_major_update - Rolls back problematic updates
create_github_pr - Creates PRs using MCP
create_github_issue - Creates issues using MCP
parse_error_for_dependency - AI-powered error analysis

4. Dependency Operations (Helper Module)

Utility functions for dependency manipulation (dependency_operations.py):

Applies updates to various dependency file formats
Rolls back specific package updates
Categorizes updates (major/minor/patch)
Finds latest versions within major releases

Functions:

apply_all_updates() - Applies all updates to dependency files
rollback_major_update() - Rolls back specific package versions
parse_error_for_dependency() - AI analysis of build errors
categorize_updates() - Categorizes updates by type
get_latest_version_for_major() - Finds latest version in major release

📦 Installation

Prerequisites

Python 3.8 or higher
Git
Node.js and npm (for checking npm packages)
pip (for checking Python packages)
Other package managers as needed (cargo, go, etc.)

Setup

Clone this repository:

git clone https://github.com/codeWithUtkarsh/AiAgentToolCalling.git
cd AiAgentToolCalling

Install Python dependencies:

pip install -r requirements.txt

Set up your Anthropic API key:

export ANTHROPIC_API_KEY='your-api-key-here'

Or create a .env file (copy from .env.example):

cp .env.example .env
# Edit .env and add your API key

GitHub MCP Setup (Required for PR/Issue Creation)

The system uses GitHub MCP (Model Context Protocol) to create Pull Requests and Issues automatically.

Prerequisites

Container Runtime (Docker, OrbStack, Podman, etc.)
- macOS: Install OrbStack (recommended) or Docker Desktop
- Windows/Linux: Install Docker Desktop
GitHub Personal Access Token
- Create at: https://github.com/settings/tokens
- Required scopes: repo, workflow

Setup Steps

1. Install Container Runtime (if not already installed)

macOS (choose one):

# Option 1: OrbStack (recommended - lightweight and fast)
brew install orbstack

# Option 2: Docker Desktop
# Download from https://www.docker.com/products/docker-desktop

2. Fix PATH for macOS/OrbStack Users

If you're on macOS and encounter "docker: command not found" errors in Python:

# Add to your shell configuration
echo 'export PATH="/usr/local/bin:/opt/homebrew/bin:$PATH"' >> ~/.zshrc
source ~/.zshrc

For bash users:

echo 'export PATH="/usr/local/bin:/opt/homebrew/bin:$PATH"' >> ~/.bash_profile
source ~/.bash_profile

3. Set GitHub Token

export GITHUB_PERSONAL_ACCESS_TOKEN='your_github_token_here'

To make it permanent:

# For zsh (macOS default)
echo 'export GITHUB_PERSONAL_ACCESS_TOKEN="your_token"' >> ~/.zshrc
source ~/.zshrc

# For bash
echo 'export GITHUB_PERSONAL_ACCESS_TOKEN="your_token"' >> ~/.bash_profile
source ~/.bash_profile

4. Verify Setup

Test your GitHub MCP setup:

python diagnose_github_mcp.py

Expected output:

✅ Container runtime: PASS
✅ GitHub token: PASS
✅ MCP connection: PASS
✅ MCP tool call: PASS

If any tests fail, the diagnostic tool will show specific instructions to fix the issue.

Supported Container Runtimes

The system automatically detects and works with:

Docker Desktop - Official Docker
OrbStack - Lightweight Docker alternative for macOS
Podman Desktop - Daemonless container engine
Rancher Desktop - Kubernetes + containers

No configuration needed - it auto-detects which one you have installed!

🚀 Usage

Automated Update with Testing (New! Recommended)

The fully automated system that updates dependencies, tests them, and creates PRs:

python auto_update_dependencies.py <repository>

Examples:

# Using full URL
python auto_update_dependencies.py https://github.com/expressjs/express

# Using owner/repo format
python auto_update_dependencies.py expressjs/express

What it does:

📊 Clones and analyzes your repository
🔄 Updates all dependencies to latest (including major versions)
🧪 Runs build and test commands
🔙 If tests fail: identifies problematic packages and rolls back major updates
✅ Creates a Pull Request if successful
🔴 Creates an Issue if updates can't be applied safely

Prerequisites:

Docker installed and running: docker --version
GitHub Personal Access Token set: export GITHUB_PERSONAL_ACCESS_TOKEN='your_token'
Git push access to the repository
Package manager tools installed (npm, pip, cargo, etc.)

📊 Sample Workflows

Workflow 1: Successful Update with Rollback

Repository: myapp (Node.js project)

📊 Analysis found 10 outdated packages:
  - express: 4.17.0 → 5.0.0 (MAJOR)
  - lodash: 4.17.20 → 4.17.21 (PATCH)
  - react: 17.0.0 → 18.2.0 (MAJOR)
  - axios: 0.21.0 → 1.6.0 (MAJOR)
  ... 6 more

🔄 Applying all updates...
✅ Updated package.json

🧪 Testing updates...
  ❌ npm test failed

🔍 Analyzing error...
  Identified: React 18 breaking change in test utilities

🔙 Rolling back React 18 → 17...
  Finding latest React 17.x: 17.0.2
  ✅ Rolled back to react@17.0.2

🧪 Testing again...
  ✅ npm install - success
  ✅ npm run build - success
  ✅ npm test - success

✅ Creating Pull Request...
  Branch: deps/auto-update-20250126
  PR: https://github.com/owner/myapp/pull/123

RESULT:
✅ Successfully updated 10 dependencies!
  - Applied 9 updates to latest versions
  - Rolled back React 18.2.0 → 17.0.2 (breaking changes)
  - All tests passing

📝 PR Summary:
  - express 4.17.0 → 5.0.0 ✅
  - lodash 4.17.20 → 4.17.21 ✅
  - react 17.0.0 → 17.0.2 (rolled back from 18.2.0)
  - axios 0.21.0 → 1.6.0 ✅
  - ... 6 more ✅

Workflow 2: Failed Update (Issue Created)

Repository: legacy-app (Python project)

📊 Analysis found 5 outdated packages:
  - django: 2.2 → 4.2 (MAJOR)
  - requests: 2.25.0 → 2.31.0 (MINOR)
  ... 3 more

🔄 Applying all updates...
✅ Updated requirements.txt

🧪 Testing updates...
  ❌ pytest failed

🔍 Analyzing error...
  Identified: Django 4.x breaking changes in models

🔙 Rolling back Django 4.2 → 2.2...
  Finding latest Django 2.x: 2.2.28
  ✅ Rolled back to Django 2.2.28

🧪 Testing again...
  ❌ pytest still failing

🔍 Analyzing error...
  Identified: Compatibility issues with Python version

🔴 Cannot apply updates safely

📋 Creating Issue...
  Issue: https://github.com/owner/legacy-app/issues/45

RESULT:
❌ Updates could not be applied safely

Issue created with details:
  - Attempted updates to 5 packages
  - Django major update causes breaking changes
  - Python version compatibility issues detected
  - Manual review and migration needed

📊 Sample Output (Legacy Mode)

Orchestrator Agent Output

================================================================================
🤖 Dependency Update Agent
================================================================================

📦 Repository: expressjs/express
🔗 URL: https://github.com/expressjs/express

📊 Running dependency analyzer on https://github.com/expressjs/express...

> Entering new AgentExecutor chain...
Cloning repository...
Repository cloned successfully

Detecting package managers...
Found: npm (package.json)

Checking outdated packages...
Found 8 outdated npm packages

================================================================================
✅ FINAL REPORT
================================================================================

# 🔄 Dependency Updates for expressjs/express

## 📦 Updated Dependencies

### ⚠️ Major Updates
- 🔴 **body-parser**: `1.19.0` → `2.0.0` (MAJOR - may have breaking changes)

### Minor Updates
- 🟡 **cookie**: `0.4.1` → `0.5.0` (minor)
- 🟡 **debug**: `2.6.9` → `2.7.0` (minor)

### Patch Updates
- 🟢 **accepts**: `1.3.7` → `1.3.8` (patch)
- 🟢 **etag**: `1.8.1` → `1.8.2` (patch)

## 🧪 Testing Instructions

Please run the following commands to verify the updates:

```bash
# Install dependencies
npm install

# Run tests
npm test

# Run build
npm run build

# Check for issues
npm run lint

⚠️ Important Notes

⚠️ This PR includes MAJOR version updates
Review changelogs for breaking changes
Run the full test suite before merging
Check for deprecation warnings
Verify build succeeds
Review any peer dependency warnings

📊 Total dependencies updated: 8 🤖 This PR was generated by the Dependency Update Agent


## 🔄 Workflow

The orchestrator agent follows this workflow:

1. **Analyze Dependencies**
   - Clone the repository
   - Detect package managers
   - Identify outdated dependencies
   - Generate analysis report

2. **Update Dependency Files**
   - Read current dependency files
   - Update version numbers
   - Preserve file formatting
   - Determine testing strategy

3. **Create PR Description**
   - Categorize updates (major/minor/patch)
   - Add testing instructions
   - Include warnings for breaking changes
   - Provide checklist

4. **Report Results**
   - Summary of updates
   - PR description ready to use
   - Next steps for the user

## 🛠️ Extending the System

### Adding New Package Manager Support

1. **Update `dependency_analyzer.py`:**

Add detection in `detect_package_manager` tool:
```python
if os.path.exists(os.path.join(repo_path, "your-config-file")):
    package_managers["your-pm"] = {
        "files": ["your-config-file"],
        "lock_files": []
    }

Create a checking tool:

@tool
def check_yourpm_outdated(repo_path: str) -> str:
    """Check for outdated packages in your package manager."""
    # Implementation here
    pass

Update dependency_operations.py:

Add update logic:

def apply_yourpm_updates(file_path: str, updates: list) -> bool:
    """Update your package manager config file."""
    # Implementation here
    pass

Update smart_dependency_updater.py:

Add testing strategy in detect_build_command tool:

# Add detection for your package manager
if package_manager == "your-pm":
    return {
        "install": "your-pm install",
        "build": "your-pm build",
        "test": "your-pm test"
    }

🔍 Agent Communication Flow

User Input (repo URL)
    ↓
Auto Update Orchestrator (auto_update_dependencies.py)
    ↓
    ├─ Check Prerequisites (Docker, GitHub Token)
    ↓
    ├─→ Dependency Analyzer Agent
    │   ├─→ clone_repository
    │   ├─→ detect_package_manager
    │   ├─→ check_npm_outdated / check_pip_outdated
    │   └─→ cleanup_repository
    │   ↓
    │   Returns: Analysis Report (outdated packages)
    ↓
    ├─→ Smart Dependency Updater Agent
    │   ├─→ detect_build_command (auto-detect test commands)
    │   ├─→ apply_updates (update dependency files)
    │   ├─→ test_updates (run build/test)
    │   ├─ If tests fail:
    │   │   ├─→ parse_error_for_dependency (AI error analysis)
    │   │   ├─→ rollback_major_update (rollback problematic package)
    │   │   └─→ test_updates (retry, max 3 attempts)
    │   ├─ If tests pass:
    │   │   └─→ create_github_pr (via Docker MCP)
    │   ├─ If tests still fail after rollbacks:
    │   │   └─→ create_github_issue (via Docker MCP)
    │   ↓
    │   Returns: PR URL or Issue URL
    ↓
Returns to User: Success (PR created) or Failure (Issue created)

🔧 Using Different LLM Providers

The system currently uses Anthropic's Claude, but you can switch to OpenAI:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4-turbo-preview", temperature=0)

Update all three agent files and set your OpenAI API key:

export OPENAI_API_KEY='your-api-key-here'

🗂️ Caching

The system includes smart caching to improve performance and reduce API calls:

How Caching Works

Repository Caching: Cloned repositories are cached locally to avoid re-cloning
Analysis Caching: Dependency analysis results are cached
Outdated Package Caching: Results from package registry checks are cached
TTL-Based Expiration: Cache expires after a configurable time period

Configuration

Set cache expiry time in hours via environment variable:

# Set in .env file or export
CACHE_EXPIRY_HOURS=24  # Default: 24 hours

Cache Location

Cache is stored at: ~/.cache/ai-dependency-updater/

Cache Management

# View cache statistics
python repository_cache.py stats

# Clean up expired cache entries
python repository_cache.py cleanup

# Clear all cache
python repository_cache.py clear

Cache Benefits

⚡ Faster repeated analyses - No need to re-clone repositories
💰 Reduced API calls - Cached package registry lookups
🌐 Works offline - Can analyze previously cached repositories
📊 Smart invalidation - Automatic expiration based on TTL

⚠️ Limitations

Some package managers require additional tools (e.g., cargo-outdated for Rust)
Large repositories may take time to clone and analyze (first time only, then cached)
Some checks require the package manager to be installed locally
Network connectivity required for cloning and checking updates (unless using cache)
Requires container runtime (Docker/OrbStack/Podman) for GitHub MCP integration
Automatically creates PRs on success and Issues when updates fail

📝 Example Scenarios

Scenario 1: Update a Node.js Project

python auto_update_dependencies.py facebook/react

Scenario 2: Update a Python Project

python auto_update_dependencies.py https://github.com/pallets/flask

Scenario 3: Analyze and Update with Auto-Rollback

python dependency_analyzer.py https://github.com/rust-lang/cargo

🤝 Contributing

Contributions are welcome! Areas for improvement:

Add more package manager support
Implement actual GitHub PR creation
Add support for monorepos
Improve version parsing and semver handling
Add caching for faster repeated analyses
Create a web interface

Please feel free to submit a Pull Request.

📄 License

MIT License - see LICENSE file for details

📚 References

🆘 Troubleshooting

"Module not found" errors

Make sure all dependencies are installed:

pip install -r requirements.txt

API Key errors

Ensure your Anthropic API key is set:

export ANTHROPIC_API_KEY='your-key-here'

npm/pip not found

Install the required package managers:

npm: Install Node.js from https://nodejs.org/
pip: Included with Python 3

Repository cloning fails

Check your internet connection
Ensure you have git installed
Verify the repository URL is correct and public

Built with ❤️ using LangChain and Claude

Name		Name	Last commit message	Last commit date
Latest commit History 47 Commits
src		src
tests		tests
.env.example		.env.example
.gitignore		.gitignore
API_README.md		API_README.md
Dockerfile		Dockerfile
GITHUB_MCP_SETUP.md		GITHUB_MCP_SETUP.md
README.md		README.md
docker-compose.yml		docker-compose.yml
pyproject.toml		pyproject.toml

Folders and files

Latest commit

History

Repository files navigation

AI Agent Tool Calling - Automated Dependency Update System

🌟 Features

Core Capabilities

Language Support

Smart Features

🏗️ Architecture

Agent Hierarchy

Complete Workflow Diagram

1. Auto Update Orchestrator (Main Entry Point)

2. Dependency Analyzer Agent (Analysis)

3. Smart Dependency Updater Agent (Update & Test)

4. Dependency Operations (Helper Module)

📦 Installation

Prerequisites

Setup

GitHub MCP Setup (Required for PR/Issue Creation)

Prerequisites

Setup Steps

Supported Container Runtimes

🚀 Usage

Automated Update with Testing (New! Recommended)

📊 Sample Workflows

Workflow 1: Successful Update with Rollback

Workflow 2: Failed Update (Issue Created)

📊 Sample Output (Legacy Mode)

Orchestrator Agent Output

⚠️ Important Notes

🔍 Agent Communication Flow

🔧 Using Different LLM Providers

🗂️ Caching

How Caching Works

Configuration

Cache Location

Cache Management

Cache Benefits

⚠️ Limitations

📝 Example Scenarios

Scenario 1: Update a Node.js Project

Scenario 2: Update a Python Project

Scenario 3: Analyze and Update with Auto-Rollback

🤝 Contributing

📄 License

📚 References

🆘 Troubleshooting

"Module not found" errors

API Key errors

npm/pip not found

Repository cloning fails

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages